THU258 Development Of A Clinical Dataset For Artificial Intelligence: A 30-Year Journey

Disclosure: S.R. Gumpeny: None. L. Gumpeny: None. Deep learning requires clean and machine-readable data for analysis. Currently there is a mismatch between the availability of accessible clinical data and the development of predictive modeling employing artificial intelligence. The scarcity is particularly acute in India. We describe the construction and evolution of electronic medical records (EMR) in an Endocrine Centre in India. December 1992 marked the end of paper documents at EDC, India. Ever since all records were stored as electronic medical records (EMR). Digitization was preceded by an analysis of work flow to ensure harmonious interaction between patient encounter and capture of data. Over years the EMR evolved across databases and was refined to improve ease of use. Rule based diagnoses, prescriptions, diet advice and auto-calculations were incorporated. Eventually it was set up on a networked platform to allow biographic, lab and anthropometric data to be entered before the patient came to the consultation chamber. Currently, the database has 90,000 individual records, including follow up data of patients with different endocrine disorders. We are at the cusp of linking the EMR with wearable sensors to make it more comprehensive. It is ripe to serve as a platform for deep learning, predictive modelling and artificial intelligence. Account sof its construction, as well as the potential for using AI in clinical care have been documented References: (1). Sridhar GR. Expanding scope of information technology in clinical care. IGI Global. DOI: 10.4018/978-1-13479-3.ch131. (2) Sridhar GR, Lakshmi G. Artificial intelligence in medicine: diabetes as a model. In (Eds) Srinivasa KG et al. Artificial Intelligence for Information management: A healthcare perspective. Studies in Big Data 88. Doi.org/10.1007/978-981-16-0415-7_14 Presentation: Thursday, June 15, 2023